Covid 'experts' are killing evidence based medicine
Eric Topol's plug of a misleading study is the latest blow
Evidence based medicine was never perfect, but it taught us important lessons. While it is often reasonable to act in a crisis or dire moment without randomized evidence, there is no excuse for failing to generate evidence year after year.
Observational research, while great for prognosis, natural history, and risk classification, is ill-suited to determine if medical interventions work. These lessons were hard fought. In the wake of WHI and CAST and ESCAPE, they were widely known. The pandemic has taken them away.
Enter Eric Topol, a cardiologist at Scripps. Topol is frequently quoted by media outlets as if he is a covid ‘expert.’ Here he touts an observational study of Paxlovid.
Of course, the study is riddled with problems. First, look at Day 0, the curves already split, a sure sign of immortal bias. But there are other problems. Michael Barnett points out that the EHR may have poorly captured vaccine status, and as such adjusting from a variable with data errors may be insufficient. Walid Gelad notes that the majority of people did not have a COVID test date, and statisticians note that the authors made bizzare imputation assumptions for those who stayed at home that ensured said immortal time.
Suffice it to the say, the paper is a train wreck. But worse, anyone with basic critical appraisal skills would know that. Yet, it was widely broadcast from a popular twitter account. Why?
If you tweet low quality studies that support the gamble, you are ingratiating yourself to the White House. That actually does pay dividends. The current White House Czar got his job because his tweets attracted the attention of cable TV, and cable TV attracted the attention of the White House. At the time, he was in contact with the White House and praised decisions nearly immediately after they occurred. From the NY Times.
What about Evidence Based Medicine?
EBM is clear on the issue of Paxlovid. We should run a randomized trial in vaccinated people, probably several, and before we spend $10 billion. We need trials in patients that look like the real Americans who are being given the pill. We should pay for the drug only insofar as the trials support (unvaccinated/ no prior covid/ & RFs - so far), and not beyond that.
Only randomized evidence is helpful like the UK Panoramic study, and non-randomized EHR fishing expeditions with immortal time bias you can see from Mars are not. Non randomized evidence will forever be plagued by differences in people who get pax and those who do not (likely being rich, health literate and well connected gets you more pax!). Confounding by indication.
Eric Topol is sadly performing a disservice to evidence based medicine, and this e.g. is just the latest among several misinterpretations (see papers on long covid he has highlighted). Many lack proper controls, and should not be published let alone highlighted, as we show in another recent paper.
The White House ran ahead of evidence. Now we have a situation of confirmation bias on steroids.
The Death of EBM
Masking, boosters in perpetuity, and Paxlovid mark the death of EBM. The USA ran zero RCTs of masking, and the CDC director testified this week before Congress that we lacked equipoise for such studies. That was categorically false. I explain in this video.
We have people defending the EUA of Paxlovid in vaccinated people without RCT data, and others who are willing to boost a 20 year old man (forever) who had COVID twice! without RCT data.
EBM is in crisis. The people who benefit are corporations and opportunists, and the people who are hurt are the American people and the last few scientists with integrity.